Fast checklist
If a forecast surprises you, run these quick checks before assuming it’s wrong:
Is all sales history imported? (recent channel connection?)
Are you looking at the right horizon & history window in the product forecast view?
Are bundles mapped to components so sub-SKUs inherit bundle demand?
Did you recently run a promo / launch that isn’t yet reflected in learned patterns?
Are you filtered to a single channel while comparing to total demand?
How Rewize forecasts (why inputs matter)
Our V2 engine blends two “experts”:
A detail model (machine learning) that reacts to recent sales, product attributes, channels, and calendar events (holidays, promos).
A trend + seasonality model that captures longer-run patterns.
We blend by horizon (near term leans on details; long term leans on trend) and reconcile across channels so channel forecasts add up to the total. If any input stream is thin, late, or mis-mapped, downstream demand can skew.
Common causes of off demand & what to do
1. Missing or partial sales history
New connection? Imports may still be running. Refresh later; if data is still light, ping us. See Why does the sync process take so long?
2. Wrong history / horizon selection
In Product details → Forecast, confirm the history window (e.g., 30d vs 1y) and future horizon you selected. Short history on a seasonal item = misleading projection.
3. Bundle demand not flowing to components
If bundles aren’t mapped, sales sit on the bundle SKU and components under-forecast. Map bundles in Bundles (Edit → + Add component). See How can I add and edit bundles?
4. Channel scope mismatch
The forecast view can include/exclude channels. Make sure the channel filters match what you’re comparing against in your reports. See Which sales channels do you support?
5. Recent promotions / anomalies
Sharp one-off spikes (e.g., Black Friday) may not fully carry forward unless patterns repeat or you extend history to include similar events. Consider manual overrides in your PO planning window.
6. Newly launched or low-volume SKUs
With little data, the model borrows broader category / peer patterns but variance is higher. Treat early forecasts as directional; recheck after a few weeks of sales.
7. Data quality issues (duplicates, wrong product mappings)
If a product exists under multiple IDs across channels or was merged incorrectly, the forecast may split demand. Verify mapping in Products & Catalog and in your connected source.
Improving forecast quality
Let initial sync finish and confirm sales depth.
Keep calendar events (major promos) flagged in your upstream data where possible.
Regularly review Bundles and ignored SKUs so demand lands on the right items.
Maintain clean channel → location fulfillment links in onboarding; it drives where demand is allocated.
Still off?
Contact us: SKU ID(s) and date range you’re reviewing. We’ll inspect the underlying data and model inputs.